Critique of Pure Reason
the Dialectic
Immanuel Kant
Copyright ©2010–2015 All rights reserved. Jonathan Bennett
[Brackets] enclose editorial explanations. Small ·dots· enclose material that has been added, but can be read as though it
were part of the original text. Occasional •bullets, and also indenting of passages that are not quotations, are meant as aids to
grasping the structure of a sentence or a thought. Each four-point ellipsis . . . . indicates the omission of a brief passage that
seems to present more difficulty than it is worth. Longer omissions will be reported between square brackets in normal-sized
type. This version follows (B) the second edition of the Critique, though it also includes the (A) first-edition version of the
Paralogisms of Pure Reason. Undecorated marginal numerals refer to page-numbers in B; ones with an ‘A’ in front refer to
A, and are given only for passages that don’t also occur in B. The likes of ..356 in the margin mean that B356 (or whatever)
started during the immediately preceding passage that has been omitted or only described between square brackets. These
marginal numerals can help you to connect this version with other translations, with the original German, and with references
in the secondary literature. Cross-references to other parts of this work include the word ‘page(s)’, and refer to numbers at the
top-right corner of each page.—The Transcendental logic divides into the Transcendental analytic, which started on page 45,
and the Transcendental dialectic, which starts here.
First launched: January 2008
Critique. . . Dialectic
Immanuel Kant
Contents
Introduction
155
1. Transcendental illusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
2. Pure reason as the seat of transcendental illusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Book 1: The concepts of pure reason
163
1. The ideas in general . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
2. The transcendental ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
3. System of the transcendental ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Book 2: The dialectical inferences of pure reason
Chapter I, The paralogisms of pure reason (first edition)
First paralogism: Substantiality . . . . . . . . . . . . . . .
Second paralogism: Simplicity . . . . . . . . . . . . . . . .
Third paralogism: Personhood . . . . . . . . . . . . . . . .
Fourth paralogism: ideality (in regard to outer relation) .
174
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
175
. 178
. 179
. 182
. . 184
The paralogisms of pure reason (second edition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Chapter 2: The antinomy of pure reason
1. System of cosmological ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2. Antithetic of pure reason . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
First antinomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Second antinomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Third antinomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fourth antinomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3. What’s at stake for reason in these conflicts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4. The transcendental problems of pure reason, considered as downright having to be soluble . . . . . . . .
5. A sceptical look at the cosmological questions raised by the four transcendental ideas . . . . . . . . . . .
6. Transcendental idealism as the key to sorting out the cosmological dialectic . . . . . . . . . . . . . . . . .
7. Critical solution of reason’s cosmological conflict with itself . . . . . . . . . . . . . . . . . . . . . . . . . . .
8. Applying the regulative principle of pure reason to the cosmological ideas . . . . . . . . . . . . . . . . . .
9. Putting the regulative principle of reason to work empirically, in connection with the cosmological ideas
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
206
. . 207
. . 211
. 213
. 216
. 219
. 222
. 225
. 230
. 233
. 235
. 238
. 242
. . 244
Critique. . . Dialectic
Immanuel Kant
Chapter 3: The ideal of pure reason
1. The ideal in general . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2. The transcendental ideal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3. Speculative reason’s arguments for the existence of a supreme being . . . . . . . .
4. There can’t be a successful ontological argument for the existence of God . . . . . .
5. There can’t be a successful cosmological argument for the existence of God . . . .
6. There can’t be a successful physico-theological argument ·for the existence of God·
7. Critique of all theology based on speculative principles of reason . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
263
. 263
. . 264
. 269
. 272
. 276
. 283
. . 287
Appendix to the transcendental dialectic
292
1. The regulative use of the ideas of pure reason . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
2. The final purpose of the natural dialectic of human reason . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
Critique. . . Dialectic
Immanuel Kant
Introduction
influenced from outside itself. . . . But the senses and the understanding are our only sources of knowledge; so error must
be brought about by ·the understanding’s being influenced
from outside itself—specifically, by· the unobserved influence
of sensibility on the understanding. What happens is that
the subjective grounds of the judgment join forces with the 351
objective grounds, making the objective grounds deviate
from their true function.1 Analogously, a moving body would
keep moving for ever along the same straight line if nothing
interfered with it, but it swerves away from that line if
another force acts on it in a different direction. To distinguish
the proper action of the understanding from the ·external·
force that is mixed in with it, we have to •regard its erroneous
judgment as the diagonal between two forces—forces that
push the judgment in different directions that enclose an
angle (so to speak)—and to •break this composite action
down into the simple actions of the understanding and of the
sensibility. In the case of pure a priori judgments this task
is performed by transcendental reflection, through which,
as I have already shown [see the explanation on page 144], every
representation is given its place in the corresponding faculty
of knowledge, so that the understanding’s influence can be
distinguished from that of sensibility.
I’m not concerned here with empirical illusions (e.g. optical illusions) that occur in the empirical use of rules of under- 352
standing that are otherwise correct, and through which the
faculty of judgment is misled by the influence of imagination.
My topic is transcendental illusion, which exerts its influence
on principles that aren’t even meant for use in experience
1. Transcendental illusion
..349
[Kant is about to warn us not to think that a ‘logic of illusion’ is a
‘doctrine of probability’. The warning looks more apt in German than
it does in English. The word standardly translated as ‘illusion’ is Schein,
cognate to the verb scheinen = ‘seem’. And the German for ‘probability’
I have already characterized dialectic in general as a logic of illusion [see page 45]. ·I
should head off right away two possible misunderstandings
of that·. (1) I don’t mean that it’s a doctrine of probability. For
probability is truth. It’s admittedly truth that is known on
insufficient grounds, so that the knowledge of it is imperfect;
but that doesn’t mean that it is deceptive; so probability
theory belongs ·not here in the dialectic but· in the analytic
part of logic. [A reminder: in this version, ‘know’ translates erkennen,
is Wahrscheinlichkeit = ‘true-seemingness’.]
which doesn’t imply anything of the sort ‘known for sure’ or ‘known
350
Introduction
through overwhelming evidence’; see the note on pages 2–3.] (2) It’s
even more wrong to identify illusion with appearance. The
essence of illusion is that it leads to error, so the concept
of illusion belongs only in contexts where ‘true’ and ‘false’
are in play. Now, there’s no work for true/false to do in
connection with •intuitions, as distinct from •judgments that
are made on the basis of intuitions. (That’s why it is right to
say that the senses don’t err—not because they always judge
correctly but because they don’t judge at all!) The domain of
operation of the concepts of truth/falsity/ error/illusion is
that of the judgment—that is, the relation of the object to our
understanding. A representation of the senses never involves
error, because it never makes any judgment whatsoever;
and there is no error, either, in any item of knowledge
that completely accords with the laws of understanding—a
natural force can’t deviate from its own laws unless it is
1
155
When sensibility is subordinated to understanding, as providing the
understanding with something to work on, it is the source of real
items of knowledge. But when that same sensibility influences how
the understanding works and dictates the judgments that it makes,
it’s the basis of error.
Critique. . . Dialectic
Immanuel Kant
Introduction
·relevant· logical rule. When we look more carefully at the
(if they were, we would at least have a criterion of their
correctness). Transcendental illusion defies all the warnings
of criticism and sweeps us out beyond the empirical use
of categories and fobs us off with a merely deceptive extension of pure understanding. I shall label as immanent any
principles that are usable only within the limits of possible
experience, and I’ll label as transcendent any principles that
profess to go beyond these limits. [‘Immanent’ comes from Latin
given case, the illusion—click!—vanishes. •Transcendental
illusion, on the other hand, persists even after it has been
detected and clearly revealed as invalid by transcendental
criticism; an example of this is the illusion in the proposition:
The world must have a beginning in time. [Kant will return to this
on page 213. •In this version, ‘criticism’ = ‘critique’. A note on page 12
The cause
of this kind of illusion is the fact that certain basic •rules
and maxims that tell us how to use our reason (subjectively
regarded as a faculty of human knowledge) have all the
appearance of being •objective principles. ·That is, rules that
really mean things of the form ‘When engaged in cognitive
activities, do X’ have the appearance of being of the form ‘The
real world is Y’. It’s not just that they are really subjective
yet appear to be objective; but also they are really advice or
commands, yet they appear to be informative propositions·.
Some ways of connecting our concepts are advantageous
to the understanding, and are ·in that sense· subjectively
necessary ·for us·, and we see these as objectively necessary
·for the world ·, i.e. as statements about what things must
be like in themselves. This is an illusion that can no more 354
be prevented than. . . .an astronomer can prevent the moon
from appearing larger as it rises, even though he knows that
it isn’t really.
So the transcendental dialectic will content itself with
exposing the illusion of transcendent judgments, while also
keeping us from being deceived by it. It can’t make the
illusion actually disappear (as logical illusion does), because
what we have here is a •natural and inevitable illusion, trading on subjective principles that it foists on us as objective;
whereas logical dialectic in exposing deceptive inferences
has to deal merely with a failure to follow the rules—i.e.
with an illusion •artificially created by something imitating
explains the choice of which word to use in a given context.]
meaning ‘remain inside’, and ‘transcendent’ comes from Latin meaning
Don’t confuse transcendent with transcendental—they are not equivalent terms. On the one hand we
have:
•the transcendental use or misuse of the categories,
which is merely an error of the faculty of judgment
when it isn’t properly reined in by criticism, so that
it doesn’t pay enough heed to the boundary of the
territory in which (the only territory in which) pure
understanding is allowed to run free.
·I said ‘use or misuse’, but in fact it is always a misuse·. The
principles of pure understanding, which I have expounded
·in the Analytic of Principles [pages 89–154]·, are for empirical
353 use only, and not for transcendental use, i.e. use extending
beyond the limits of experience. In contrast to that we have
•transcendent principles—actual principles that encourage us to tear down all those boundary-fences,
step across them, and claim an entirely new domain
that recognises no limits of demarcation.
If my critique can succeed in exposing the illusion in
these alleged principles, then the principles that are for
merely empirical use can be set off against the others by
being called immanent principles of pure understanding.
•Logical illusion—the illusion that a formally invalid argument is valid—is a mere imitation of the form of reason, and
it happens only when we don’t attend carefully enough to the
‘climb over’.]
156
Critique. . . Dialectic
Immanuel Kant
a valid inference. So there we have it: there’s a natural
and unavoidable dialectic of pure reason. It’s not something
that a bungler might get tangled in through ignorance, or
something that a sophist has contrived so as to confuse
thinking people. Rather, it is inseparable from human
reason; even after its deceptiveness has been exposed, it
355 will go on playing tricks with reason, continually tricking it
into momentary aberrations that have to be corrected over
and over again.
Introduction
longer than the original, but adds nothing to what he meant.]
We are faced, then, with a division of reason into (1) reasonused-logically and (2) reason-used-transcendentally, and
now we have to hunt for a higher concept of reason that
has (1) and (2) as special cases of it—i.e. the more general
concept of reason that is an ingredient in both of the more
specific concepts of (1) and (2). (This is the ‘difficulty’ to
which I have referred.) To characterize the higher or most
general concept of reason, we need to assemble a table or
chart setting out all the concepts that fall under it; and a
clue to doing that is provided by what we found with the
faculty of understanding. That too has both logical and
transcendental uses, and it turned out that the logical uses
provide the key to the whole story of the understanding,
including its transcendental part. Just think back or look
back to the way we moved from •the table of judgment-kinds
[page 49] to •the table of categories [page 52] and from that to
•the table of principles of pure understanding [page 99]. Well,
I shall show that a disciplined account of the basic logical
ways in which reason can be used will point to an over-all
account of the nature and shape of reason as a whole—the
genealogical tree of the concepts of reason. In doing this I’ll
be taking reason to be the faculty of principles. (This is in
contrast to the understanding, which I have been treating as
the faculty of rules. I have sometimes spoken of ‘principles
of the understanding’; I’ll explain that shortly.)
2. Pure reason as the seat of transcendental illusion
A. Reason in general
All our knowledge starts with the •senses, moves up from
there to the •understanding, and ends with •reason—our
highest faculty—so that it can work up the materials provided by intuition, bringing them under the highest unity
of thought. As I set myself to explain this highest cognitive
faculty, I find myself in some difficulty. Like the understanding, reason can be used (1) in a merely formal (i.e.
logical) manner, in which it abstracts from all content of
knowledge. But reason also has (2) a real use, because it
contains within itself the source of certain concepts and
principles that it doesn’t borrow either from the senses or
from the understanding. For a long time now logicians have
defined our ability to use reason in the (1) formal way as
our ability or faculty for making mediate inferences. ·i.e.
for drawing conclusions from two or more premises·;. . . .but
this doesn’t throw any light on the (2) other use of reason, in
which reason itself gives birth to concepts.
[We are about to encounter talk about syllogisms. Any argument with
the form of this:
(1) Some bullies are cowards,
(2) All cowards are depressed, therefore
(3) Some bullies are depressed
is a syllogism. Its major premise is (2), because it contains the predicate
of the conclusion (‘depressed’); and (1) is the minor premise.]
The term ‘principle’ is ambiguous. It is often used to stand ..356
for any item of knowledge that can be used as a principle,
[Kant’s next two sentences are hard to understand unless one knows
what is to come later. The present version of this paragraph is much
157
Critique. . . Dialectic
Immanuel Kant
even if its origin doesn’t qualify it as being a principle. Any
universal proposition, even one derived from experience
through induction, can serve as the major premise in a
syllogism, ·which involves its being ‘used as a principle’
in the weak sense of being used as a basis from which to
infer something else·, but that doesn’t mean that it is a
principle ·properly so-called·. Mathematical axioms such
as ‘There can be only one straight line between two points’
are instances of universal a priori knowledge, and therefore
relate as ‘principles’ to all their instances. (·And because
they are known a priori they are more like ‘principles’ strictly
so-called than are empirically established propositions such
as ‘All cowards are depressed’·.) But this doesn’t entitle us
to say that this property of straight lines is something that
357 we know from principles. In fact we know it only through
pure intuition [see page 30].
[Kant now offers an obscure paragraph whose main point
is that because •any ‘All. . . ’ proposition can be the major
premise in a syllogism, and so can in that way be used as a
principle, therefore •the a priori propositions associated with
the understanding ‘can be called “principles” in the sense
that they can be used as principles’. Then:]
But if we consider those basic propositions of pure understanding in the light ·not of how they can be used but
rather· of what their source is, then we can see that they are
nothing like principles ·in the strict sense of propositions·
expressing knowledge based on concepts. Our ability to have
the a priori knowledge they express comes from ·two sources,
neither of which consists in concepts, namely·
•pure intuition (for the mathematical propositions) or
•conditions that have to be satisfied for any experience
to be possible (·for the others·).
·Consider a proposition of the second type·: Every event has
a cause. This can’t be inferred merely from the concept of
Introduction
event; on the contrary, it’s a basic proposition that points to
the conditions that enable us to have a determinate event
concept in the first place. [In this passage and some later ones,
‘event’ translates a German phrase meaning ‘thing that happens’. That’s
what events are—things that happen.]
Thus, the understanding can’t supply us with any •synthetic
items of knowledge derived from concepts; and •those are the
only things that I call ‘principles’ period; though any univer- 358
sal proposition—·and especially any that can be known a
priori ·—can be described as ‘relating to such-and-such in a
principle-like way’.
It has long been wished. . . .that instead of the endless
complexity of the laws of the land we could find their principles, because that’s our only hope of ‘simplifying’ the law.
·There’s nothing •problematic about the thought that there
are such principles, because· the laws we are considering
here are only constraints that we have imposed on our
freedom, which means that ·if they are harmonised and
simplified under very general legal principles, the latter·
are directed to something that is entirely our own work—
something that we have generated out of our own ·legal·
concepts. Contrast that with the thought that objects in
themselves, the very nature of things, stand under principles,
and are determined according to mere concepts. ·That is
•problematic·: it seems impossible, or at least quite contrary
to common sense. Well, we’ll look into that in due course.
My present point is just that
•knowledge derived from principles, strictly and properly so-called
is something quite different from
•knowledge obtained merely through the understanding,
though the latter can be principle-like in being more basic
than some other knowledge. . . .
359
158
Critique. . . Dialectic
Immanuel Kant
Introduction
judgment is added, ·then the inference of (2) from(1) is
non-immediate·. I call immediate inferences inferences of
the understanding, and non-immediate ones inferences of
reason. The proposition ‘All men are mortal’ contains the
propositions ‘Some men are mortal’, ‘Some mortal beings
are men’, and ‘Nothing that isn’t mortal is a man’—so these
are all immediate conclusions from it, ·each drawable in an
inference of the understanding·. On the other hand, the
proposition (1) ‘All men are mortal’ doesn’t contain (2) ‘All
learned beings are mortal’ (it doesn’t involve any use of the
concept of learned being), so (2) can only be inferred from (1)
only with the help of a mediating judgment.
Just as understanding can be seen as the faculty that
uses rules to unify appearances, reason can be seen as the
faculty that uses principles to unify the understanding’s
rules. Thus, reason never applies directly to experience or
to any object. What it applies to is the understanding: its
role is to give an a priori unity by means of concepts to the
understanding’s complex web of items of knowledge. This
‘unity of reason’, as we may call it, is nothing like the unity
that the understanding can create ·by bringing appearances
under its concepts·.
That’s the best I can do to explain the general concept
of the faculty of reason—or the best I can do without using
examples. They will be provided later on.
[Warning: Kant’s use here of ‘mediating judgment’ (Zwischenurteil, betweenjudgment) is misleading. If he were using it properly, he would be talking
B. The logical use of reason
about the case where to get from P to R you have to get from P to Q and
then from Q to R. (That would make the need for a mediating judgment
360
We distinguish •what is immediately known (e.g. A figure
bounded by three straight lines has three angles) from •what
is only inferred (e.g. The sum of those three angles equals
two right angles). We’re constantly in need of inferences,
and eventually we get used to inferring—so much so that we
stop being aware of the difference between immediate and
inferred knowledge, and often treat as being immediately
perceived what has really only been inferred. . . . In every
process of reasoning there is
a subjective matter: although •I can’t see that P ⇒ R except by bringing
(1) a fundamental proposition (the premise), and
(2) another proposition (the conclusion that is drawn
from (1)), and finally
the inference (logical sequence) by which the truth (2) is
inseparably connected with the truth of (1).
that needs to be understood: the standard German word for ‘syllogism’
in Q, •you are smart enough to see that P ⇒ R without getting help from
Q.) Anyway, that is not what Kant means when he speaks of mediating
judgments. His real topic here is simply cases where R doesn’t follow
from P alone but does follow from P together with Q. His announced
theory really is that what is logically special about reason is that it is used
in inferring conclusions from pairs of premises. That isn’t a load-bearing
part of what’s important in the Dialectic, but it figures in some of Kant’s
preliminary moves, so we need to get straight about it.—Something else
is Vernunftschluss = ‘inference of reason’. Some of what Kant says about
such inferences really does fit syllogisms and only syllogisms (e.g. the
technical term ‘major premise’), but much of the time he is talking more
broadly about inferences-from-pairs-of-premises. From now on this version will usually translate Vernunftschluss by ‘inference of reason’. In the
If (2) is contained in (1) in such a way that it can be derived
from (1) without the mediation of a third proposition, the
inference is called immediate. But if (2) can’t be reached
from the item of knowledge contained in (1) until another
following paragraph, the schematic S-M-P example, which isn’t Kant’s, is
tied to one very simple and basic kind of syllogism, narrowly so-called.
But you’ll soon see that his topic is broader than that.]
159
Critique. . . Dialectic
Immanuel Kant
In every inference of reason I first think a rule (the major
premise) through the understanding:
All M are P.
Secondly, I bring a known item ·S· under the condition ·M·
of the rule by means of judgment (the minor premise):
All S are M.
Finally, I determine [here = ‘establish some fact about’] the known
item ·S· by applying the predicate ·P· of the rule:
All S are P,
361
arriving at this—the conclusion—a priori through reason.
There are different kinds of inference of reason—three of
them, in fact, corresponding to the three kinds of judgments.
They are:
(1) categorical,
(2) hypothetical,
(3) disjunctive.
How do we decide which category a given inference of reason
belongs to? By looking at the form of its major premise, i.e. at
how that premise relates the two items that it involves. [The
Introduction
it is the parts-of-a-logical-division relation expressed in the
proposition R or S.
In most cases the judgment that forms the conclusion is
set as a problem—to see whether it follows from judgments
already given, ones through which a quite different object is
thought. [The element M in the categorical case, P in the hypothetical
case, and R in the disjunctive case is ‘quite different from’ anything in
the conclusion.] I look in the understanding to see how this
conclusion is situated there; I’m trying to discover whether
it stands under certain conditions according to a universal
rule. If I find such a condition ·embodied in a rule·, and if the
conclusion relates to it in the right way, then the conclusion
is deduced from the rule—which is also valid for other objects
of knowledge. We see here that in inference reason tries
to reduce the complex web of knowledge obtained through
the understanding to the smallest number of principles
(universal conditions) thereby bringing it into the highest
possible unity.
C. The pure use of reason
rest of this paragraph is an addition to what Kant wrote, but it consists
In a (1) categorical
inference of reason like the one semi-illustrated above, it is
the subject-predicate relation expressed in the proposition
that all M are P. In a (2) hypothetical one, of the form:
If P then Q,
P,
therefore Q,
it is the ground-consequent relation expressed in the proposition that if P then Q. And in a (3) disjunctive inference, of
the form
only of borrowings from things he will say later.]
[In this paragraph, Kant speaks of the Vergleichung of one proposition
with another, standardly translated as ‘comparison’. It’s hard to avoid
that, but what he really means here is ‘comparing’ not in the sense of
likening P to Q but only in the sense of laying them side by side so as to
Here are two prima 362
facie possible accounts of the basic status of reason:
(1) Reason can be considered all on its own; it is an
independent source of concepts and judgments that
come from it alone and give it a relation to objects.
(2) Reason is a merely subordinate faculty, whose role
is to impose a certain ‘logical’ form on given items of
knowledge. It is through reason that things known by
means of the understanding are determinately related
take in their inter-relationship in a single thought.]
R or S,
Not-R,
therefore S,
160
Critique. . . Dialectic
Immanuel Kant
to one another, with lower items of knowledge being
brought under higher ones,. . . .this being done by
comparing them.
Which of these is right? That’s the preliminary question we
are now facing up to (·you’ll see in a moment why I call it
‘preliminary’·). ·The answer is (2) rather than (1). Reason is
perhaps not ‘merely subordinate’, in that· it demands that
the multiplicity of rules ·of the understanding· be unified by
principles ·of reason·. (In doing this work,
•reason makes the ·output of· the understanding hang
together in a thoroughly connected whole, by bringing
it under principles,
just as
•the understanding connects up the various outputs
of intuition, by bringing them under concepts.)
But a principle of reason doesn’t prescribe any law for objects;
it doesn’t contain anything that is needed as a basis for
knowing objects or knowing anything about them (·that
being what enables the understanding to prescribe laws
for objects·). Reason is merely a subjective law for the
orderly management—the housekeeping—of our stock of
understanding-outputs,. . . .aiming at the greatest possible
economy in our use of them. It doesn’t entitle us to demand
363 that the objects have a uniformity that will make things
easier for our understanding and increase its reach; so we
can’t ascribe any objective validity to the maxim ·in which
reason demands that the output of the understanding be
unified·. ·With the preliminary question thus answered, we
now come to the big question that will be with us for a long
time·. In a word, the question is: Does reason in itself—i.e.
does pure reason—contain a priori synthetic principles and
rules, and what might such principles consist in?
If pure reason is capable of a transcendental principle
through which it yields synthetic knowledge, what will it
Introduction
be based on? We get sufficient guidance in answering that
from ·two points about· the formal and logical procedure of
reason.
First, an inference of reason doesn’t concern itself with
•intuitions, aiming to bring them under rules (as the understanding does with its categories). What it deals with are
•concepts and •judgments. Thus, even if pure reason does
·somehow· concern itself with objects, what it is immediately
related to are not •objects and the intuition of them, but
rather •the understanding and its judgments, which do deal
at first hand with the senses and their intuition for the
purpose of establishing facts about their object. The unity
of understanding is the unity of a possible experience, but
the unity of reason is nothing like that. The proposition
Every event has a cause contributes to making the unity of
experience possible; it’s because of this making-experiencepossible that the understanding can use its concepts to
pull experience together through synthetic propositions like
that one. In contrast with that, reason doesn’t have the job
of making experience possible; and that deprives it of any 364
chance of imposing on experience any such synthetic unity
as is imposed by Every event has a cause.
Secondly, when reason is put to use logically, it starts
with some judgment and tries to find a universal rule of
which the judgment is a special case. (The universal rule
is the major premise, and the judgment in question is the
conclusion.) In doing this, reason is acting on its maxim: ·
When you have an item of knowledge, find something
more general of which it is a special case,
or, to say the same thing in more technical terms·,
•When you have an item of knowledge, find the condition by which it is conditioned.
Now, the major premise of any inference-of-reason also
falls within the scope of reason’s seek-the-condition maxim,
161
Critique. . . Dialectic
Immanuel Kant
which means that reason tells us to look for something still
more general from which it follows. That involves going from
the inference—
P0 , Q, therefore R—
to a prior inference of reason whose conclusion is P0 and
whose major premise is some proposition P−1 ; and from that
to a still earlier one whose conclusion is P−1 , and so on
·backwards and upwards·. All of this happens in accordance
with reason’s principle—its very own principle—
Given any conditioned item of knowledge obtained
through the understanding, find the unconditioned
whereby the understanding can be completely unified.
We may want to treat this logical maxim as a principle of
pure reason—i.e. ·to regard it not merely as •a command that
tells us what to do, but as •a statement saying that something is the case—namely that for everything conditioned
there is a condition·. This would involve us in assuming
that if something conditioned is given, the whole series of
conditions. . . .is likewise given, i.e. is contained in the object
and its connection. (Notice that if this whole series exists, it
is itself unconditioned.)
Such a principle of pure reason is obviously synthetic;
something that’s conditioned is analytically related to •some
condition but not to •the unconditioned. And other synthetic
propositions must follow from it—propositions of which pure
365 understanding knows nothing, because it deals only with
items that are conditioned. [These days we might say that the
unconditioned doesn’t appear on the understanding’s radar screen.] If
there actually is anything unconditioned, we’ll have to pay
special attention to all •the features of it that distinguish it
from everything that is conditioned, and •they’ll provide the
raw material for many synthetic a priori propositions.
Any principles arising from this supreme principle of pure
reason will be transcendent in relation to all appearances, i.e.
The concepts of pure reason
there can’t be any adequate empirical use for the supreme
principle. (So it will be entirely different from all principles of
understanding, because their use is wholly immanent—·they
don’t transcend experience, because· their only theme is the
possibility of experience.) Consider the principle:
•The series of conditions extends to the unconditioned.
(This might be offered either as telling us a truth about
•what there is in the world out there or as making a demand
about •how we are to behave when we think about the world
out there.) Does this principle have objective applicability?
What does it imply concerning the empirical use of the
understanding? Or is there no such objectively valid principle
of reason, but only a logical command that instructs us to
advance towards completeness by working our way up to ever
higher conditions, thereby giving our knowledge the greatest
possible unity of reason? Might it be the case that this
command should never have been viewed as a transcendent 366
principle of pure reason, and that we went too fast when we
postulated that the objects out there in the world actually
include an unrestrictedly complete series of conditions? And
if that is how things stand, what other misunderstandings
and delusions may have crept into the inferences of reason
·that we conduct·?. . . . Answering these questions will be
my task in the transcendental dialectic, which I’m aiming to
develop from its deeply concealed sources in human reason.
I’ll shall divide the Dialectic into: •Book 1 [pages 163–173] on
the transcendent concepts of pure reason, •Book 2 [pages 174–
291] on its transcendent and dialectical inferences, and •an
Appendix [page 292—315].
162
Critique. . . Dialectic
Immanuel Kant
Book 1:
The concepts of pure reason
transcendental ‘ideas’. I’ll now explain and justify this label.
The concepts of pure reason
1. The ideas in general
In advance of settling whether there can be any concepts
derived from pure reason, we know this much: if we can get
them, it will be through inferences and not through reflection. Concepts of understanding are also thought a priori,
367 antecedently to experience and for the sake of experience;
but all they give is the unity of reflection on appearances
that have to belong to a possible empirical consciousness.
Those concepts make it possible for us to have knowledge
and to settle facts about the objective world. And they don’t
come to us through inferences; ·there are indeed two reasons
why they couldn’t possibly do so·. (1) They first provide the
material required for making inferences, ·so we can’t do any
inferring until we have those concepts·. (2) ·There’s nothing
we could infer them from ·: they aren’t preceded by any a
priori concepts of objects from which they could be inferred.
Their objective reality isn’t based on anything inferential;
its sole basis is the fact that they constitute the intellectual
form of all experience; so it must always be possible to show
their application in experience.
The label ‘concept of reason’ tells us from the outset that
we’re dealing here with something that can’t be confined
within experience, because it concerns a body of knowledge
of which any empirical knowledge is only a part—indeed it
may be that the whole of possible experience. . . .is only a
part of it. No actual experience is ever completely adequate
to it, yet every actual experience belongs to it. Concepts of
reason enable us to conceive, and concepts of understanding
enable us to understand. . . . Just as I have labelled the
pure concepts of understanding ‘categories’, so I shall give
the concepts of pure reason a new name, calling them
Despite the great wealth of our languages, a thinker often
finds himself at a loss for the expression that exactly fits
his concept, and this lack prevents him from being really
intelligible to others or even to himself. He could coin a
new word, but that amounts to claiming to legislate for 369
language—and you can’t often get away with that! Before
trying that way out—which is always a long shot—we should
scout around in a dead learned language, to see if it provides
both the concept and a suitable word for it. Even if those
who first launched the word were a bit careless, so that the
use of it was somewhat wobbly even back then, it’s always
better •to latch onto the meaning that distinctively belongs
to it (whether or not we’re sure that it was originally used in
precisely this sense) than •to defeat our purpose by making
ourselves unintelligible.
When we want to distinguish a certain concept from related ones, and there’s just one word whose existing meaning
exactly fits it, we would be wise to use that word sparingly—
keeping it to its own proper meaning and not also using
it, for stylistic variety, as a synonym for other expressions.
Otherwise we may stop focussing intently on that word,
mixing it up with lots of other words whose meanings are
quite different; and if that happens, we’ll lose also the
•thought that the •word expresses and could have preserved. 370
From the way Plato used the term ‘idea’ we can see that
he meant it to stand for something that not only •couldn’t
be borrowed from the senses but even •extends far beyond
the concepts of understanding (which Aristotle was busy
with)—because nothing that fits it can ever be encountered
in experience. Plato held that ideas are archetypes of the
163
Critique. . . Dialectic
Immanuel Kant
things themselves [= ‘models from which things are copied’], unlike
the categories, which are merely keys to possible experiences.
In his view, ideas issued from highest reason, through which
human reason comes to share in them; but our reason is
no longer in its original state, and has to strain to recall the
old now-obscure ideas, by a process of recollection (which
is called philosophy’). I’m not going to conduct a textual
enquiry into what this great philosopher meant by ‘idea’.
I merely remark that it isn’t at all unusual to find. . . .that
we understand an author better than he has understood
himself. Not having pinned down his concept exactly enough,
an author’s intention is sometimes belied by what he has
said, or even by what he has thought.
Plato knew very well that •our faculty of knowledge feels
a need for something much higher than merely spelling out
371 appearances according to a synthetic unity so as to be able
to read them as experience; and •that our reason naturally
soars to items of knowledge that have to be recognised as
having their own reality rather than being mere fictions of
the brain, despite the fact that they go far beyond the bounds
of experience—so far that no empirical object can ever fit
them.
Plato found the chief instances of his ideas in the field
of the practical [here = ‘moral’], i.e. in what rests on freedom,
which is the subject of items of knowledge that are produced
only by reason.2 If you try to derive the concepts of virtue
from experience. . . . you’ll turn virtue into something that
varies according to time and circumstance, a slippery nonentity that can’t be brought under any rule. We’re all well
2
The concepts of pure reason
aware that the truth is nothing like that. We know that
if someone is held up as a ‘perfect example of virtue’, we 372
judge this by comparing the person, the alleged ‘perfect
sample’, with the true original that we have in our minds.
This original is the idea of virtue. Objects of experience
can serve as ·approximate· examples of it (showing that
proofs that what the concept of reason commands is at
least somewhat feasible), but they can’t serve as ·perfect·
archetypes of virtue. [This uses the word (Urbild) that was translated
as ‘archetype’ two paragraphs back; but here, and from now on, Kant
thinks of an Urbild not as a model from which other things are copied,
but rather as a model or ideal example to which we may approximate.]
None of us will ever act in a way that matches up to what is
contained in the pure idea of virtue, but that doesn’t prove
that there’s something chimerical about this thought. It’s
only by means of this idea that we can make any judgment
as to moral worth or unworth; so the idea serves as an
indispensable basis for any approach to moral perfection—
even if we don’t get very close because of the obstacles in
human nature. . . .
Plato’s Republic has become proverbial as a striking example of the kind of dreamy perfection that could only exist in
an idle thinker’s brain, and he has been ridiculed for claiming
•that a monarch can’t rule well unless he participates in the
ideas. [Here as everywhere in this half of the Critique, ‘idea’ is used
only as a Platonic or Kantian technical term.] We would, however,
be better advised to run with •this thought, and, where the
great philosopher leaves us without help, to shed light on
it through our own efforts rather than discarding it on the
wretched and harmful pretext that it isn’t practicable.
373
It’s true that he also extended his concept of idea so as to cover
items of speculative knowledge, provided that they were pure and
given completely a priori. He even extended it to mathematics, although what that science is about can be found only in possible
experience. I can’t go with him down that road. . . .
A constitution providing for the greatest possible human freedom under laws that make the freedom of
each consistent with the freedom of everyone else
164
Critique. . . Dialectic
Immanuel Kant
—that is, to put it mildly, a necessary idea; it must be made
basic not only in the initial design of a constitution but
also in all its laws. (I state the idea in terms of •freedom,
not of •the greatest happiness, for happiness will take care
of itself if freedom is assured.) In drafting a constitution,
we must initially abstract from the present obstacles. ·Let’s
think a little about the nature of these obstacles to successful
government·. Rather than being inevitable upshots of human
nature, perhaps they arise rather from something that could
be remedied, namely the neglect of genuine ideas in making
laws. Legislators commonly ·excuse their failures by· appealing to ‘adverse experience’—·i.e. to contingent circumstances
that thwarted their plans·. Actually, nothing could be more
harmful or more unworthy of a philosopher than that. The
‘adverse experience’ wouldn’t have occurred if at the right
time those institutions had been set up in accordance with
ideas, rather than the ideas being displaced by crude conceptions which, just because they were derived from experience,
nullified all ·the legislators’· good intentions. The more
legislation and government are brought into harmony with
the above idea [i.e. the one indented earlier in this paragraph], the
less punishment there would be; so it was quite reasonable
for Plato to maintain that in a perfectly structured state
no punishments would be needed. It may be that this
perfect state won’t ever come into being; but that doesn’t
374 stop the idea from being valid. What it does is to set this
maximum—·‘the greatest possible human freedom’·—before
us as an archetype, something we can move towards, so as
to bring the legal organisation of mankind ever nearer to its
greatest possible perfection. How far can we go along that
line? How big a gap must there be between the ·archetypal·
idea and what we actually achieve? No-one can answer this,
and no-one should try, because this is all about freedom,
which can pass beyond any specified limit.
The concepts of pure reason
Plato saw clear proofs that ideas have an explanatory
role not only •in the moral sphere, where human reason
exhibits genuine causality so that ideas are working causes
of actions and their outputs, but also •in regard to nature
itself. A plant, an animal, the orderly arrangement of the
cosmos—presumably therefore the entire natural world—
clearly show that they are possible only according to ideas.
No individual creature coincides ·exactly· with the idea of
what is most perfect in its kind; just as no human being
coincides ·exactly· with the idea of humanity, though each
of us carries that idea in his soul as the archetype of his
actions. Despite this, these ideas are completely determinate
unchangeable individuals in the Supreme Understanding ·of
God·, and they are the ultimate causes of things.
[Kant offers a guarded expression of approval for Plato’s
appeal to ideas outside the moral sphere. Then:] But where ..375
Plato’s doctrine renders a very special •service is in connection with the principles of morality, legislation, and religion,
where the experience (of the good) is itself made possible
only by the ideas—incomplete as their empirical expression
must always remain. This •service hasn’t been recognized,
because it has been judged in accordance with empirical
rules—the very things that Plato’s approach has shown can’t
validly be treated as principles ·in our moral thinking·. When
we are studying nature, experience supplies the rules and
is the source of truth; but when it comes to the moral laws,
experience is (alas!) the mother of illusion! It is very bad
behaviour to derive laws prescribing what I ought to do from
what is done, or to limit laws on that basis.
Following out these considerations is what gives philosophy its own special dignity; but just now we must occupy
ourselves with a less grand but still worthwhile task, namely
levelling the ground and making it firm enough to support 376
these majestic moral edifices. ·Why does it need to be made
165
Critique. . . Dialectic
Immanuel Kant
firm?· Because this ground has been honeycombed by subterranean workings that reason, in its confident but fruitless
search for hidden treasures, has carried out in all directions.
What we have to do now is to get some insight into the
transcendental use of pure reason, its principles and ideas,
so that we can be in a position •to get the facts about what
influence pure reason has and •to make a judgment as to its
value. [Kant pleads with the philosophically serious reader
to use ‘idea’ only in its original meaning rather than using it
as a label for ‘any and every species of representation’. There
are plenty of terms for each kind of representation, he says,
and he gives a list—‘a chart’—of terms with their definitions.]
The concepts of pure reason
Fifth level:
·Pure concepts can be schematised, i.e. amplified by
something sensible·.
A pure concept originating solely in the understanding,
with no input from sensibility, is a notion.
·And so at last we rise to our present topic, which involves
‘notion’ but seems not to come from any two-part division of
notions, namely:
Sixth level·:
A concept that is formed from notions and outruns
the possibility of experience is an idea.
Anyone who has familiarised himself with these distinctions
must wince when he hears the representation of the colour
red called an ‘idea’. It oughtn’t even to be called a concept of
understanding, a notion.
[•In this version, each bold-type item is the one that re-appears at the
next level up. •In this one case Erkenntnis is translated as ‘cognition’,
because the generally preferred ‘item of knowledge’ sounds too peculiar.
See note on pages 2–3. •Despite its prominence here, this is the last we
hear of ‘notion’ as a technical term.]
2. The transcendental ideas
Bottom level:
The genus is ‘representation’.
When this is accompanied by consciousness it is
perception.
Second level:
Perception considered merely as a state of the person
is ‘sensation’.
Perception considered as perception of something is
cognition.
Third level:
A cognition relating directly to an individual object is
an ‘intuition’.
A cognition relating indirectly to objects, through
features that many objects may share, is a concept
Fourth level:
Empirical concepts.
Pure concepts.
The Transcendental Analytic gave us an example of how
the mere logical form of our knowledge can give rise to
pure a priori concepts which represent objects prior to all
experience. (Strictly speaking, rather than •representing
objects they •indicate the synthetic unity without which we
couldn’t have empirical knowledge of objects.) The differ- 378
ent forms of judgment. . . .generated categories that direct
all our use of understanding in experience. In the same
way we can expect that the different forms of inferences
of reason. . . .will generate special a priori concepts (we can
call them ‘pure concepts of reason’ or ‘transcendental ideas’)
which will determine how understanding is used in dealing
with experience as a totality.
The function of reason in its inferences is to give ·greater·
universality to items of knowledge. . . . Consider the proposition, ‘Caius is mortal’. I could get this from experience
166
Critique. . . Dialectic
Immanuel Kant
by means of the understanding alone, ·leaving reason out
of it·. But I am after ·something more general; I’m looking
for· a concept (in this case, the concept man) that contains
the condition under which the predicate. . . .of this judgment
(‘is mortal’) is given; and after I have brought the predicate
under this condition taken across its whole range (‘All men
are mortal’), I proceed on that basis to settle on the item of
knowledge about my object (‘Caius is mortal’).
Accordingly, in the conclusion of an inference of reason
379 we restrict a predicate to a certain object, having first thought
it in the major premise under a given condition taken across
the whole range of that condition. This fact about the size
of the range is called universality or totality. . . . So the
transcendental concept of reason is nothing but
•the concept of the totality of the conditions for any
given conditioned item.
What makes possible the totality of conditions is the unconditioned, and conversely the totality of conditions is always
itself unconditioned. [This use of ‘conversely’ here suggests that
The concepts of pure reason
expressed by (1) ‘S is M’, (2) ‘If P then Q’, and (3) ‘R or S’·.
So we have to look for three kinds of unconditioned item:
(1) the categorical synthesis in a subject; (2) the hypothetical
synthesis of the members of a series; (3) the disjunctive
synthesis of the parts in a system.
So there are exactly three kinds of inference of reason,
each of which moves up through prosyllogisms to the relevant unconditioned item: (1) to the subject that is never
itself a predicate; (2) to the presupposition that doesn’t 380
presupposes anything further; (3) to an aggregate of the
members of the division of a concept such that nothing
further is needed to complete the division. So the pure
concepts of reason—
concepts of totality in the synthesis of conditions,
·i.e. concepts of going the whole way in looking for a
condition for every conditioned item·
—are necessary at least as setting us the task of extending
the unity of understanding, where possible, right up to the
unconditioned. They are based on the nature of human
reason, ·which is essentially committed to the demand for
conditions·. It may be that there isn’t anything for these
transcendental concepts actually to apply to; in which case
the only good they do is to direct the understanding in such a
way that when it is extended to the uttermost it is completely
free of inconsistency.
Kant meant to say that something involves a totality of conditions if
and only if it involves something unconditioned. But that isn’t what
So we can give a general explanation of
what a pure concept of reason is—·i.e. what an idea is·—by
saying that it’s a concept of something unconditioned, when
the concept is thought of as a basis for the synthesis of
the conditioned. ·That means: as a basis for a process of
connecting conditioned items with one another; for example,
•discovering a causal chain among certain events
would be
•conducting a synthesis of (causally) conditioned items,
and similarly with the other relevant relations·. There will
be exactly as many •pure concepts of reason as there are
•kinds of relation that the understanding represents to itself
by means of the categories. ·There are just three of these,
he actually says.]
·T HE RIGHT WAY TO USE ‘ ABSOLUTE ’·
While I’m dealing with ‘the totality of conditions’ and ‘the
unconditioned’ as equivalent labels for all concepts of reason,
I come on another expression (·as well as ‘idea’·) that I
can’t do without but can’t safely use, because long-standing
misuse has made it ambiguous. The word is ‘absolute’. Like
just a few others, this word in its original meaning was fitted
to a concept that no other word in the language exactly
suits. So if •the word is lost, or if (same thing) it is used
167
Critique. . . Dialectic
Immanuel Kant
with several different meanings, •the concept itself will be
lost too. This is a concept that reason is very busy with,
and giving it up would do great harm to all judgments in
transcendental philosophy. (1) The word ‘absolute’ is now
often used merely to indicate that something is true of a
thing considered in itself, and therefore true of its inward
nature; in this sense, ‘x is absolutely possible’ means that
x is in itself possible—which is the least that could be
said about it. (2) But the word is also sometimes used
to indicate that something holds true in all respects, without
limitation (e.g. absolute despotism), and in this sense ‘x is
absolutely possible would mean that x is in every relation
(in all respects) possible—which is the most that can be
said of x’s possibility. ·From here on, though I shall be
discussing both senses of ‘absolute’, I shall use the word only
in sense (2), reserving ‘intrinsic’ for sense (1)·. Sometimes a
statement is true in both senses of ‘absolute’: if something
is (1) intrinsically impossible then it is (2) impossible in
any relation, and therefore absolutely impossible. But in
most cases the two meanings are infinitely far apart: if
something is (1) in itself possible, we can’t conclude that
it is also (2) possible in every relation, and thus absolutely
possible. We’ll see later on that absolute necessity doesn’t
always depend on intrinsic necessity, and therefore shouldn’t
be treated as equivalent. If the opposite of something is
382 intrinsically impossible, this opposite is of course impossible
in all respects, and the thing itself is therefore absolutely
necessary. But we can’t run this inference the other way,
arguing that if something is absolutely necessary its opposite
is intrinsically impossible, i.e. that the absolute necessity of
things is an intrinsic necessity. . . . The loss of a concept that
is of great importance for speculative philosophy must matter
to you if you are a philosopher. [In Kant’s usage, ‘speculative’ is
381
The concepts of pure reason
of theories’, and doesn’t carry any of the sense of ‘guesswork’ that the
I hope, then, that it will matter to you that we
should pin down and carefully preserve the word on which
the concept depends.
So there it is: I shall use the word ‘absolutely’ in contrast
to what holds only comparatively, i.e. in some particular
respect; referring to what is valid without restriction in
contrast to what is restricted by conditions. [As well as ab-
word has today.]
solut, which he is discussing here, Kant often uses schlechthin, which
means ‘without qualification’. It could often be translated by ‘absolutely’,
and in previous translations it often is; but the present version will
use ‘absolute(ly) only for absolut, and translate schlechthin by ‘utterly’
or ‘unqualifiedly’ or some such expression. When Kant contrasts (1)
things that are principle-like in this or that way with (2) things that
are schlechthin Prinzipien, he is translated on page 158 as contrasting
(1) with ‘principles period’. Grossly unhistorical, but it does capture his
meaning.]
Now a transcendental concept of reason always aims at
absolute totality in the synthesis of conditions, and its only
terminus is in what is unqualifiedly unconditioned, i.e. is
not conditioned in any respect. For pure reason leaves to
the understanding everything that kicks off from the objects 383
of intuition, or rather from the synthesis of such things in
the imagination. Reason’s only concern is with absolute
totality in the use of the concepts of the understanding; it
takes the synthetic unity that is thought in the category and
tries to track it up to something unqualifiedly unconditioned.
We can call this the unity of reason in appearances, and
that expressed by the category the unity of understanding.
Reason isn’t concerned with the understanding considered
as containing the ground of possible experience. Why?
Because •no experience is unconditioned, so •the concept
of the absolute totality of conditions isn’t applicable in any
experience, ·so •reason has nothing to do or say down at that
the opposite of ‘practical’ or ‘moral’; it means ‘having to do with the truth
168
Critique. . . Dialectic
Immanuel Kant
level·. But reason is concerned with the understanding in
another way: it tells the understanding what direction to take
towards a certain unity, of which the understanding itself
has no concept. What unity? It’s the unity that would come
from uniting all the acts of the understanding, in respect of
every object, into an absolute whole. The objective use of the
pure concepts of •reason is, therefore, always transcendent,
while that of the pure concepts of •understanding must
always be immanent, because the only way to use them is in
application to possible experience.
[Kant now has a paragraph in which he repeats what he
has already said •about the ‘transcendental’ nature of pure
concepts of reason, •about their role as direction-setters for
the understanding, and •about the ideas of practical reason
as having a larger and more direct role in human life than
do those of speculative reason, the latter being our concern
..385 in this book. Then:]
In saying (as we must) that the transcendental concepts
of reason are only ideas, we aren’t taking them to be superfluous and empty. Although they can’t latch onto any
object, they can in a basic and unnoticed way be useful to
the understanding as a canon for its extended and consistent
use [re ‘canon’, see note on page 25]. What this provides for the
understanding is not
•more knowledge than it would have by means of its
own concepts ·unguided by reason·,
but rather
•better and more extensive guidance for the acquiring
of knowledge.
386 Not to mention the fact that concepts of reason may enable
us to move across from thoughts about •nature to thoughts
about •morality. . . . I’ll deal with that in a later work. In this
work our concern is. . . .only with reason in its speculative
use—and indeed, more narrowly, with its transcendental
The concepts of pure reason
speculative use. Let’s take a tip from our procedure in the
deduction of the categories, by considering the logical form
of knowledge through reason. . . .
Reason. . . .is the faculty of inferring, i.e. judging mediately
(by bringing the condition of a possible judgment under the
condition of a given judgment). The given judgment is the
universal rule (major premise)
•·All men are mortal·.
What brings the condition of another possible judgment
under the condition of the rule is the minor premise
•·Caius is a man·.
The judgment which applies the predicate of the rule (·‘mortal’·)
to the brought-under case ·of Caius· is the conclusion
•Caius is mortal.
387
The rule says of some predicate that it applies to everything
that satisfies a certain condition. That condition (·mortality·)
is found to be satisfied in an actual case (·Caius·). What has
been asserted to be universally valid under that condition
is therefore to be regarded as valid also in the present
case, which satisfies that condition. It’s easy to see what is
happening here: reason is arriving at an item of knowledge
through acts of the understanding that constitute a series
of conditions. Here is an example, concerning my way of
arriving at the proposition that (3) All bodies are alterable. I
start from the proposition that
(1) Everything composite is alterable.
This item of knowledge is quite distant from (3); it doesn’t
involve the concept of body, though it does involve the
condition of that concept, alterability. I then proceed from
(1) to a proposition that is less remote from (3), and stands
under the condition of (1), namely the proposition that
(2) ·All· bodies are composite.
From this I finally pass to
(3) ·All· bodies are alterable,
169
Critique. . . Dialectic
Immanuel Kant
which connects the more distant item of knowledge (alterable)
with the knowledge actually before me. By this procedure
I have arrived at an item of knowledge (a conclusion) by
means of a series of conditions (the premises). [In one
fiercely compressed sentence, Kant says things that can
fairly be spelled out as follows: An inference of reason in
which we pass from conditions (in the premises) to something
conditioned (in the conclusion) can sometimes be part of a
longer series of conditions-to-conditioned inferences, going
in either direction.
•In one direction, the longer series takes the conditions in the premises of the original inference and
provide conditions of them, and then conditions of
those conditions, and so on upwards.
•In the other direction, the longer series takes the
conditioned item in the conclusion of the original
inference and provide items of which it is a condition,
and then items of which those items in turn are
conditions, and so on downwards.
This can’t happen with disjunctive inferences of reason, but
it can happen with either of the other two forms of inferences
of reason—categorical (= subject-predicate) and hypothetical
(= if-then). The most natural kind of example (Kant doesn’t
give any) of the hypothetical form of inference takes the
use of the hypothetical ‘If . . . then’ to express facts about
what causes what. We explain the fact that •Q by putting
together our knowledge that •P’s being the case would cause
Q to be the case and our knowledge that •P. Then we can
move upward into the fact that P is caused to be the case
by O, which is caused to be the case by N, and so on back
up the causal chain; or downwards into the fact that Q
causes R to be the case, which causes S to be the case, and
so on down the causal chain. Examples of an elongated
inference of reason that has the categorical form are harder
The concepts of pure reason
to provide, or even to describe; they will be returned to
[page 172]. The disjunctive form doesn’t come into this
because a disjunction doesn’t have a direction.]
..388
But we soon become aware that how the faculty of reason
works in the •ascending series of inferences of reason, in
which we infer items of knowledge by looking at
•conditions as being conditioned in their turn,
is quite different from how it behaves in the •descending
series, in which we look at
•conditioned items as being conditions in their turn.
In the ascending inference the item of knowledge is given only
as conditioned; to arrive at it by means of reason we have
to assume that all the members of ·the ascending series·,
the series on the side of the conditions, are given—·the
crucial point being that we have to think of that entire series
as already complete ·. . . . In the ·descending· series, the
one on the side of the conditioned, the one that looks at
consequences, our only thought is of a series in process of
coming into existence, not one already presupposed or given
in its completeness. . . . Thus, if an item of knowledge is
viewed as conditioned, reason is forced to regard the series
of conditions in the ascending line as completed and as
given in its totality. But if the same item of knowledge is
viewed as a condition of further items of knowledge that 389
constitute a series of consequences in a descending series,
reason doesn’t care •how far this downward series extends,
or •whether a totality of the series is possible. That’s because
reason doesn’t need any such series in order to draw its
conclusion. [Kant’s development of this point is expressed
rather technically. What it comes down to, expressed here
(though not by him) purely in terms of the causal kind of hypothetical inference of reason, is this:] Reason is compelled
to regard any present event as the upshot of all its causes;
without knowing whether that series has a first member (an
170
Critique. . . Dialectic
Immanuel Kant
uncaused cause) or rather stretches back to infinity, with
no first member, it has to regard the event as having such a
totality of causes in its ancestry; the proposition reporting
this one event can’t be counted as true unless the entire
series of its causes is unconditionally true. (This holds even
if it is admitted that we can’t possibly grasp a totality of
conditions.) Reason requires this, by announcing that its
knowledge is a priori determined as necessary, either •in
itself (in which case it needs no grounds) or else •derivatively
as a member of a series of grounds—a series which is, taken
as a whole, unconditionally true.
The concepts of pure reason
which he derives a three-part relation-based classification
(r-bc) of concepts of pure reason (i.e. transcendental ideas).
In the paragraph after this one he will say that this r-bc
coincides with the classification he has already presented
on the basis of •logical form, namely the division into (1)
categorical, (2) hypothetical, and (3) disjunctive. What really
matters is not the r-bc itself but rather a classification that
Kant supposedly derives from it, namely:]
(1) ideas containing the absolute (unconditioned) unity of
the thinking subject,
(2) ideas containing the absolute unity of the series of
conditions of appearance, and
(3) ideas containing the absolute unity of the condition of
all objects of thought in general.
The thinking subject is what (1) psychology is about, the
sum-total of all appearances (the world) is what (2) cosmology
is about, and the thing that contains the highest condition
of the possibility of all that can be thought (the Being of
all beings) is what (3) theology is about. Thus, pure reason
provides the ideas for (1) a transcendental doctrine of the
soul, a rational psychology, (2) a transcendental science of 392
the world, a rational cosmology, and (3) a transcendental
knowledge of God, a rational theology. The understanding
can’t produce even a sketch of any of these sciences—even
when it is supported by the highest logical use of reason, i.e.
by all possible inferences through which we aim to move from
given appearances right up to the most remote members of
the empirical synthesis. Each of these sciences is an entirely
pure and genuine product of pure reason—or problem of
pure reason!
How, exactly, do the pure concepts of reason come under
these three headings? I’ll answer that fully in the next
chapter, where we’ll see that they follow the guiding-thread
of the categories. ·If you are wondering how the categories,
3. System of the transcendental ideas
390
Our topic is not logical dialectic, which ignores the •content
of knowledge and confines itself to exposing the fallacies
concealed in the •form of inferences of reason. Rather, it is
a transcendental dialectic that has to contain, completely a
priori, the origin of certain items of knowledge derived from
pure reason as well as of certain inferred concepts whose objects •can’t ever be given empirically and therefore •lie wholly
beyond ·the reach of· the faculty of pure understanding. The
transcendental use of our knowledge, both in inferences
and in judgments, has a natural relationship to its logical
use; and this relation has shown us •that there can be only
three kinds of dialectical inference of reason, corresponding
to the three kinds of inference through which reason can
arrive at knowledge by means of principles, and •that in
all of these its business is to ascend from the conditioned
synthesis to the unconditioned—i.e. from something to which
the understanding always remains restricted to something
that the understanding can never reach.
[Kant now presents an obscure account of three kinds
of relation that can be involved in a representation, from
171
Critique. . . Dialectic
Immanuel Kant
which are concepts of the •understanding, come into this
story about our concepts of •reason, I’ll point out here that·
pure reason latches directly not onto objects but onto the
understanding’s concepts of objects. ·So much for the general
point, but what about the details·? I shall contend that
(1) reason, simply by the synthetic use of that very function of which it makes use in categorical inferences of
reason, is necessarily brought to the concept of the
absolute unity of the thinking subject; that
(2) the logical procedure used in hypothetical inferences
of reason leads to the ideal of the utterly unconditioned in a series of given conditions, and finally that
389
(3) the mere form of the disjunctive inference of reason
must necessarily involve the highest concept of reason,
that of a Being of all beings—a thought that, at first
sight, seems utterly paradoxical.
When I complete my account ·in the next chapter·, I shall
make clear how all that can be the case.
Strictly speaking, there can’t be an objective deduction
of these transcendental ideas, like the one I gave for the
categories, because the ideas—just because they are ideas—
don’t relate to any object in such a way that they could be
(·or, for that matter, fail to be·) true of it. But a subjective
derivation of them from the nature of our reason can be
given, and in this chapter I have given it.
The concepts of pure reason
whole range of possibilities amongst them; rather than some being made
subordinate to others, they are all treated as on a level, as somehow going
together or concurring. You’ll recognize that this is just the same 1-2-3
that we have been dealing with in the past few pages. This note makes a
feeble job of relating disjunction to ‘concurrence’, but the blame for that
may lie with Kant.] It’s easy to see that what pure reason has in
view is the absolute totality of the synthesis on the side of the
conditions (whether of inherence, of dependence, or of concurrence); it isn’t concerned with absolute completeness on
the side of the conditioned. It’s only the former that is needed
in order to presuppose the whole series of the conditions and
present it a priori to the understanding. Given a complete
(and unconditioned) condition, we don’t need any concept
of reason for the continuation of the series: every step in
the downward direction from condition to conditioned—·from 394
conditions to what they are conditions of ·—is taken by the
understanding itself. The transcendental ideas, therefore,
serve only for going up the series of conditions to the unconditioned, i.e. to principles. As regards the intellectual
journey down from conditions to the conditioned, reason
does indeed make a very extensive logical use of the laws of
understanding, but it’s not a transcendental use. If we form
an idea of the absolute totality of a synthesis in a downward
series—e.g. an idea of the whole series of all future alterations
in the world—this is a mental entity that we have chosen to
create, not something we are forced to presuppose by the
nature of our reason. . . .
Finally, we also come to realize that the transcendental
ideas themselves hang together to form a certain unity,
and that it’s by means of them that pure reason draws
all its items of knowledge together to form a system. The
advance from (1) the knowledge of oneself (the soul) to (2)
the knowledge of the world, and by means of this to (3)
the primordial being, ·God·, is so natural that it seems
[We are about to meet three technical terms that have to be understood:
(1) ‘inherence’,
(2) ‘dependence’,
(3) ‘concurrence’.
In (1) a categorical or subject-predicate proposition, some property is
said to inhere in a subject—e.g. mortality inheres in Caius. In (2) a
hypothetical proposition something is said to depend on something else—
e.g. the ball’s starting move depends on its having been hit. In (3) a
disjunctive proposition, two or more possibilities are said to divide the
172
Critique. . . Dialectic
395
Immanuel Kant
to resemble reason’s logical advance from premises to a
conclusion.3 . . . .
[The phrase ‘primordial being’ translates the German Urwesen. The
prefix Ur- is used to convey the idea of something that is the basic source
of x, the fundamental origin of all the Fs, or the like. (English has no such
resource except in words openly borrowed German—e.g. such English
words as ‘urkingdom’ and ‘urtext’.) Some Kant translators use ‘original
being’; but ‘original’ doesn’t colloquially carry the weight and solemnity
of Ur-. Thus, ‘primordial’, here and throughout; with apologies, and this
explanation.]
3
Metaphysics has only three ideas as the proper objects of its enquiries: God, freedom, and immortality—so related that the combination of God with freedom leads inevitably to immortality. Any
other matters that metaphysics may deal with are merely means
of arriving at these ·three· ideas and of establishing their reality.
Reason needs the ideas not for the purposes of natural science but
in order to pass beyond nature. Insight into them would put the
faculty of speculative reason in sole charge of theology and morals,
and, through the union of these two, likewise religion—which means
that it has sole charge of the highest ends of our existence. [Kant is
now going to use ‘analytic’ and ‘synthetic’ in a way that was quite
standard in his day but is entirely different from the senses he
has given these words up to here. In the present sense, ‘analytic’
and ‘synthetic’ are labels for two methods of presentation of some
doctrine. An •analytic presentation starts with things we all know to
be true and works its way from those to the theory or doctrine that
explains and is supported by them. A •synthetic presentation goes
in the opposite direction: it starts with the fundamental theses of the
doctrine to be expounded, and works from those to various of their
consequences, which could include the things-we-already-know that
are the starting-point for the analytic format.] In a •systematic presentation of the ideas, the synthetic order would be more suitable;
but before we get to that there has to be a •preliminary workingthrough of the materials, and for that the analytic order—the reverse
of the synthetic—is better. It lets us start from what is immediately
given us in experience—advancing from the doctrine of the soul to
the doctrine of the world and from that to knowledge of God.
173
The concepts of pure reason